LSE AI Marketing Test With Into-it: Judgment Analysis

I once watched a marketing calendar become a moral document. It looked harmless enough, a grid of campaign dates, email sends, audience segments and approvals. But every cell contained a judgment about who mattered, what tone was acceptable, which promise the brand could safely make, and when silence was better than another post. That is why the London School of Economics experiment with Into-it feels bigger than another AI productivity story. MarketingWeek reports that the London School of Economics has partnered with martech startup Into-it to create a fully autonomous AI marketing function. The phrase sounds like a dare, which is probably why the publication framed the project around whether a university can create a fully AI marketing team. But the more interesting question is not whether AI can schedule, segment, generate and optimize. It is whether marketing operations were ever as procedural as we pretended.

The Org Chart Becomes

a Workflow According to MarketingWeek, the LSE and Into-it project is explicitly aimed at a fully autonomous AI marketing function, not just a tool that helps a person write faster. That distinction matters because most AI adoption in marketing has so far been framed as assistance. The marketer remains the protagonist, and the software is a very fast intern. Agentic systems change the shape of that story. In LSE Business Review, Terence Tse describes agentic AI as artificial intelligence that proactively makes decisions and drives actions toward preset goals, assessing situations, forming plans and executing them with minimal human oversight. That is not a better autocomplete. It is a miniature operating model. Once marketing becomes an operating model run by agents, the org chart starts to blur. Planning, testing, reporting and iteration can become loops rather than meetings. The old friction of handoffs may shrink, but so can the informal checks that live inside those handoffs. A colleague asking, “Are we sure this sounds like us?” is not always documented as governance, but it often is.

Automation Does Not Remove Accountability LSE Business Review argues that

agentic AI signals a fundamental shift in how we approach work and problem solving. In marketing, that shift lands in an especially sensitive place because the work is both measurable and interpretive. A campaign can hit its targets and still feel wrong. The overlooked question is simple: if an autonomous marketing function makes a bad call, who owns the decision? The vendor that built the system, the institution that set the goal, the marketer who approved the guardrails, or the executive who wanted efficiency in the first place? Springer Nature’s Evolutionary and Institutional Economics Review describes AI as introducing triangular agency relationships and new information asymmetries, which is a formal way of saying responsibility can become harder to see just when decisions become easier to execute. That does not make the LSE experiment reckless. It makes it useful. A serious institutional test can reveal where autonomy creates leverage and where it creates ambiguity. The lesson for teams is to evaluate autonomous marketing not by asking whether it can produce output, but by asking whether its decisions can be explained, challenged and reversed.

Brand Judgment Is

the Scarce Skill LSE Executive Education notes that digital marketing is becoming more complex and AI driven, and that marketers face both greater challenges and unprecedented opportunities. It also points to the need for confidence with AI and analytics, not merely curiosity. That framing is important because the human role does not disappear when the machine gets better. It moves upstream. MIT Sloan Management Review frames a similar problem for marketing leaders: many recognize the opportunity of generative AI but struggle with where to begin, especially when weighing opportunity against manageable risk. A fully autonomous marketing function sharpens that tradeoff. The easier it becomes to automate the visible work, the more valuable the invisible work becomes. That invisible work is brand judgment. It is knowing when an efficient message is too needy, when personalization becomes intrusive, when a trend is not for you, and when the correct conversion strategy is restraint. AI can learn patterns, but institutions still need people to decide which patterns deserve power.

What To Watch Next MarketingWeek’s report on LSE and Into-it gives marketers

a concrete anchor for a debate that can otherwise drift into abstraction. The practical takeaway is not to copy the experiment wholesale. It is to map your own marketing function by decision type: what can be automated, what needs review, and what should remain a human call because it defines the brand. The next useful marketing team may not be the one with the most AI tools. It may be the one with the clearest boundary between execution and judgment. If autonomous agents can run more of the machinery, humans may need to become more explicit about values, taste and accountability than they have ever been. So before asking whether AI can run your marketing, maybe ask the more uncomfortable question: do you know which parts of your marketing should never be allowed to run themselves?